In game two Tan Zhongyi took the lead in the match by winning a “drawn” endgame over the champion, Ju Wenjun, and in game three the champ returned the favor by doing exactly the same thing to her challenger. In both games the player with White enjoyed a small but sustained initiative into the endgame, won a pawn though the position remained drawn with best play, and kept pressing until the defender cracked and lost.
The similar plot line doesn’t mean that everything was the same. Game 2 was a d-pawn opening, while the current game saw White meet a 2…e6 Sicilian with 3.c4. This version of the Maroczy Bind doesn’t seem particularly dangerous - or particularly “bind-y” for that matter, as Black achieved …d5 on move 7. (That said, I would have preferred 6…Nf6, and think Tan’s 6…Nge7 was best met by 7.Nb5.) As in game 1 the challenger found herself with an isolated d-pawn, and while she managed to saddle her opponent with an isolated c-pawn in return it came at the cost of the bishop pair.
To break her opponent’s sustained pressure Tan found a clever exchanging combination that surrendered her d-pawn to reach an ending with a pair of rooks and opposite-colored bishops. The position was objectively drawn, but as discussed above (and in my post on and comments to game 2) that evaluation doesn’t guarantee that the human sitting at the board will actually achieve the draw. It didn’t happen in round two and it didn’t happen in this game either - as you can see for yourself here (with my notes).